|Laliberte, Andrea -|
|Goforth, Mark -|
|Steele, Catriana -|
Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: November 18, 2011
Publication Date: November 22, 2011
Repository URL: http://handle.nal.usda.gov/10113/57946
Citation: Laliberte, A.S., Goforth, M., Steele, C.M., Rango, A. 2011. Multispectral remote sensing from unmanned aircraft: image processing workflows and applications for rangeland environments. Remote Sensing. 3(11):2529-2551. Interpretive Summary: Unmanned aircraft systems (UAS) are ideally suited for obtaining high spatial resolution imagery for rangeland monitoring due to the ability to deployed the UAS repeatedly over the same area. A multispectral sensor is more desirable than a digital camera operating in the visible light range for greater differentiation of vegetation species and opportunity for quantitative remote sensing. We integrated a light weight multispectral camera in a small UAS. The camera acquired ten-bit radiometric data in six narrow spectral bands ranging from the blue to the near infrard. We describe challenges and solutions associated with efficient processing of the multispectral imagery to obtain orthorectified, radiometrically calibrated image mosaics for the purpose of rangeland vegetation classification. Automated processing methods were developed for file conversion, band-to-band registration, radiometric correction, and orthorectification. We obtained high correlations between the airborne spectral reflectance and ground-based reflectance measurements for soil and vegetation targets. An image mosaic classification was derived for a 100 ha area with an overall classification accuracy of 87%. Comparison of the airborne multispectral data with imagery acquired from a WorldView-2 satellite showed good correlations, indicating the potential for using UAS-acquired imagery as ground truth for satellite imagery, or for upscaling the fine resolution data to larger areas . Our results indicate that UAS are viable platforms for obtaining quality multispectral remote sensing data for rangeland applications.
Technical Abstract: Using unmanned aircraft systems (UAS) as remote sensing platforms offers the unique ability for repeated deployment for acquisition of high temporal resolution data at very high spatial resolution. Most image acquisitions from UAS have been in the visible bands, while multispectral remote sensing applications are less common. In this paper, we describe challenges and solutions associated with efficient processing of multispectral imagery to obtain orthorectified, radiometrically calibrated image mosaics for the purpose of rangeland vegetation classification. We developed automated batch processing methods for file conversion, band-to-band registration, radiometric correction, and orthorectification. An object-based image analysis approach was used to derive a species-level vegetation classification for the image mosaic with an overall accuracy of 87%. We obtained good correlations between 1) ground and airborne spectral reflectance (R2=0.92), and 2) spectral reflectance derived from airborne and WorldView-2 satellite data (R2=0.82) for selected vegetation and soil targets. UAS-acquired multispectral imagery provides quality high resolution information for rangeland applications with the potential for upscaling the data to larger areas using high resolution satellite imagery.